Abstract
The cochlea plays a key role in the transmission from acoustic vibration to neural stimulation upon which the brain perceives the sound. A cochlear implant (CI) is an auditory prosthesis to replace the damaged cochlear hair cells to achieve acoustic-to-neural conversion. However, the CI is a very coarse bionic imitation of the normal cochlea. The highly resolved time-frequency-intensity information transmitted by the normal cochlea, which is vital to high-quality auditory perception such as speech perception in challenging environments, cannot be guaranteed by CIs. Although CI recipients with state-of-the-art commercial CI devices achieve good speech perception in quiet backgrounds, they usually suffer from poor speech perception in noisy environments. Therefore, noise suppression or speech enhancement (SE) is one of the most important technologies for CI. In this study, we introduce recent progress in deep learning (DL), mostly neural networks (NN)-based SE front ends to CI, and discuss how the hearing properties of the CI recipients could be utilized to optimize the DL-based SE. In particular, different loss functions are introduced to supervise the NN training, and a set of objective and subjective experiments is presented. Results verify that the CI recipients are more sensitive to the residual noise than the SE-induced speech distortion, which has been common knowledge in CI research. Furthermore, speech reception threshold (SRT) in noise tests demonstrates that the intelligibility of the denoised speech can be significantly improved when the NN is trained with a loss function bias to more noise suppression than that with equal attention on noise residue and speech distortion.
Highlights
A cochlear implant (CI) is an auditory prosthesis playing an essential role in restoring hearing ability for patients with severe-to-profound sensorineural hearing impairment [1, 2]
This study investigates the effect of different loss functions, i.e., the way measuring the difference between neural networks (NN) output and the target signal, on their performance on NN training
There always exists some α, the values vary at different signal-to-noise ratio (SNR), with which the weighting loss (WL)-MASK front end achieved better performance than MSE-MASK
Summary
A cochlear implant (CI) is an auditory prosthesis playing an essential role in restoring hearing ability for patients with severe-to-profound sensorineural hearing impairment [1, 2]. Xu et al [26] proposed a masking-based SE, in which the NN to estimate the masking gain was trained with a loss function containing separately computed speech distortion and residual noise. We developed a DL-based SE as a front end to the signal processing strategy of CIs. A long-short term memory (LSTM) network was trained to estimate the TF masking gains. By adjusting the weights for trading off the speech distortion and the noise residue, their contributions to speech intelligibility for CI recipients were investigated, upon which an LSTM trained with preference-biased-loss was developed. To further investigate the effect of α in trading off the speech distortion and residual noise in the electrodograms, we computed and compared the current units of the enhanced electrograms and the clean ones. The fluctuation in −5 dB might be because the network has not seen a −5 db SNR during the training
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
